Himax Emerging as a Key AI Upstream Winner in TSMC's COUPE (Silicon Photonics), Significantly Boosting Growth Visibility for 2026–2028

$Taiwan Semiconductor(TSM.US) $Himax Tech(HIMX.US)

I previously shared my prediction that Himax might be a potential supplier for TSMC.
https://longportapp.cn/zh-CN/topics/26052416

My latest supply chain survey indicates that TSMC's COUPE (silicon photonics) development and ecosystem visibility have improved markedly. Furthermore, it's confirmed that Himax is the exclusive supplier of micro-lens arrays for the first and second generations of COUPE FAUs. Below are more updates and details.

Summary & Conclusions:

Thanks to COUPE, Himax is poised for robust expansion over the next few years. I estimate Himax revenue (USD) to reach approximately $1.16B, $1.42B, and $2.4B in 2026–2028, respectively (vs. the consensus of $1.11B for 2026), and EPS around $1.00, $1.60, and $3.40 (vs. the consensus of $0.80 for 2026).
The first COUPE generation is already fully developed and undergoing mass production validation. Second-generation mass production validation should begin in 1H26.
TSMC's earnings call and NVIDIA CEO Jensen Huang's statements may seem contradictory, but both are fundamentally positive about COUPE's.
Himax is the exclusive micro-lens array supplier for the first and second COUPE generations. A second supplier is unlikely to emerge before the third generation at the earliest.
In the next six months, several events could further boost market visibility for COUPE/silicon photonics.

Detailed Updates:

1.Among TSMC's developing technologies, COUPE is receiving the highest priority, which alone provides multi-year visibility. The development timeline is as follows:

4Q24: Finalize first-gen specifications and begin sampling
2025–2H26: Mass production validation for the first generation
2H26: First-gen mass production
1H26: Finalize second-gen specifications and begin sampling
1H26–2H27: Mass production validation for the second generation
4Q27–1Q28: Second-gen mass production
The key differences in the second generation revolve around performance upgrades, improved optical coupling efficiency, and higher manufacturability.

2.TSMC's earnings call vs. Jensen Huang's statements: not really contradictory

On January 16, TSMC mentioned silicon photonics could move to mass production in about 1–1.5
years
On January 17, NVIDIA CEO Jensen Huang, during a visit to Taiwan, told local media that while NVIDA is collaborating with TSMC on silicon photonics, it will take several years.
Among TSMC's clients, AMD is likely the first to adopt COUPE, continuing its aggressive trend of
using TSMC's advanced technologies (e.g., SoIC, 2nm).
Rubin Ultra could be the first NVIDIA product to use COUPE if mass production is on time (late 2027).
The above explains why TSMC's timeline and Huang's statements do not conflict.

3.Based on my recent supply chain checks, if there are no big production problems, FOCI will be the sole FAU (ReLFACon) provider for both Gen 1 and Gen 2 COUPE, and Himax is the exclusive supplier of the micro-lens array. The earliest a second supplier might enter the market would be for the Gen 3.

4.Key assumptions for Himax's outlook include:

COUPE micro-lens array shipments of around 6 million, 50 million, and 240 million units in 2026–2028, respectively.
The micro-lens array makes up roughly 30–40% of the total FAU cost, with a gross margin of 45–50%.
Each GPU and network chip needs about 35–45 FAUs and 6–8 FAUs, respectively.
The revenue and EPS estimates assume no growth in Himax's existing businesses since 2026. Any improvement there would further strengthen overall momentum.

5.COUPE/silicon photonics potential catalysts in the next six months

Watch for announcements or insights from NVIDIA, Himax, and FOCI during upcoming earnings calls.
At Nvidia's GTC event, the company may reveal future products/technologies. Technically, to achieve a step-change in classic computing power—and better integrate with quantum computing—transitioning to silicon photonics/COUPE is necessary.